Masahiro Watanabe Atmosphere & Ocean Research Institute ...

Masahiro Watanabe
Atmosphere & Ocean Research Institute
University of Tokyo
hiro@aori.u-tokyo.ac.jp
Collaborators:
M. Yoshimori, Y. Kamae, M. Kimoto (AORI, Univ of Tokyo),
H. Shiogama, T. Yokohata, T. Ogura, S. Emori (NIES),
J. D. Annan, J. C. Haargreaves (JAMSTEC)

Did climate sensitivity get converged in CMIP5
Equilibrium climate sensitivity (ECS) to 2xCO 2 in CMIP models
IPCC report FAR SAR TAR AR4 AR5
climate sensitivity 1.5-4.5K 1.0-3.5K 1.5-4.5K 2.1-4.4K
Mitchell et al. (1990), Kattenberg et al. (1996), Cubasch et al. (2001), IPCC (2007)
ECS in 4xCO2 experiments by some of the CMIP5 models
2-4.6 K
@Crown copyright Met Office
Slide courtesy of K Taylor

Model diversity
Structural differences in the multi-model ensemble (MME)
make it difficult to understand sources of uncertainty
CMAP
Annual-mean precipitation biases in CMIP3
MME
IPCC (2007)

Different groups use CMIP MMEs differently
Non-modeling groups tend to compare ensembles (e.g.
CMIP3 vs CMIP5) w/o tracing changes in a particular
model
Modeling groups can do the same, but also trace the
changes across versions of their own model and
attribute difference in the system’s behavior to
physical processes (parameterization details)

Equilibrium climate sensitivity
Gregory plots for 4xCO2 experiments
CMIP3 version
MIROC3med(T42)
2x =3.6K
N = F − α∆T
CMIP5 version
MIROC5 (T85)
2x =2.6K
SAT
SAT
Watanabe et al. (2010, JC)
Cloud-shortwave (SWcld) feedback is the key component

Multiple timescales of the cloud feedback
MIROC3.2 MIROC5 4xCO2 150y 150y
HadGEM2-ES 4xCO2
slope in the
natural variability
slope in the
natural variability
CRE [W/m 2 ]
SST [K]
Watanabe et al. (2011, Clim. Dyn.)
+ ann. mean, single run (150y)
○ decadal mean, single run
△ ann. mean, ensemble avg (20y)
★ initial month of △
SAT [K]
Courtesy of Tim Andrews
Sign of the SWcld feedback is
determined by the fast response of
clouds during a few years in MIROCs

Fast response of low clouds
4xCO2 runs
(20y ensemble)
Regime composite of low cloud wrt 500
MIROC3
MIROC5
Low cloud amount [%]
Tropical mean < 0 Tropical mean > 0
500hPa
Watanabe et al. (2011, Clim Dyn)
CTL
CTL
C
P C ( ) d P C ( )
d
l l l
Thermodynamic driving
Tropical-mean changes in C l & SWcld are due to a subtle difference
between C l decrease / increase over the subsidence / ascent regime

ENSO and climate sensitivity
Reasons may be different in the two models, but a
common feature of strong ENSO in the control run low
ECS
MIROC5 PPE-C (JUMP2)
HadCM3 PPE-C (QUMP)
SWcld feedback [W/m 2 /K]
Nino 3.4 SST std dev [K]
Corr=-0.67
Nino 3.4 SST std dev [K]
Equilibrium climate sensitivity [K]
Shiogama et al. (2012, Clim. Dyn.) Toniazzo et al. (2008)

ENSO and climate sensitivity
Compare 10 models having
* Largest negative SWcld feedback
* Smallest negative SWcld feedback
Larger SWcld feedback
Wetter mean state over the central
equatorial Pacific & stronger ENSO
ENSO amplitude [K]
wet cold tongue dry
Watanabe et al. (2011, JC)
Shiogama et al. (2012)

Forcing-Feedback in two PPEs
MIROC3 PPE
Multi-Physics Ensemble
(MPE)
CMIP3 MME
Climate
sensitivity
MIROC5
AGCM
MIROC5 MPE
MIROC5 PPE
Structural difference > Parametric difference
Any strategy to link them
Replacing one or more
schemes in MIROC5 w/
old ones:
Std (=MIROC5)
(old)CLD
(old)CUM
(old)VDF
(old)CLD+CUM
(old)CUM+VDF
(old)CLD+VDF
(old)CLD+CUM+VDF
Watanabe et al. (2012, JC)

MPE: Filling the gap between PPEs
Climate sensitivity vs SWcld feedback
Processes are nonlinear, e.g.,
Extratropics
Global
Tropics
(old)CLD -> small impact
(old)VDF -> small impact
(old)CLD+VDF -> large impact!!
SWcld feedbacks
STD
CLD+
CUM
CLD+
VDF
Watanabe et al. (2012, JC)
‘Feedback occurs thru the interaction of a suite of parameterized
processes rather than from any single process’ (Zhang & Bretherton 2008)

What can we do next
Further analysis to the CMIP5 MME on the cloud
feedback, adjustment, radiative forcing etc.
-> will come out at the CFMIP meeting (May 2012, Paris)
Paris)
Hierarchical modeling to deepen our understanding of
the cloud feedback processes (e.g. Brient & Bony 2012)
Extensive use / coordinated analysis of the PPEs
“Seamless model evaluation” ― Jim Murphy

Similarity in Pattern
Cloud-shortwave forcing (SWcld)
in CMIP5 AMIP 4xCO2 exps
HadGEM2(+0.23W/m 2 ) IPSL CM5A (+1.07W/m 2 )
Change in the tropical
vertical motion in MIROC5
4xCO2
Ocean
CNRM (+0.85W/m 2 )
MIROC5 (+0.70W/m 2 )
CTL
Land
Courtesy of M. Webb
Roughly positive (warming) over oceans,
negative (cooling) over lands
ω [hPa/s]
Initial spin-up of the Walker cell
due only to 4xCO2

ΔSWcld in idealized experiments
L000(Aquaplanet)
Dependence of ΔSWcld on the land size
L000
L180(continent in w hemi.)
SWcld [W/m 2 ]
L060
L120
Available in CMIP5 MME
L180
L240 L300
SAT [K]
[W/m 2 ]
Land-sea contrast counteracts the cloud
decrease over the tropical oceans and
hence the strong cloud-shortwave forcing

Papers available at http://tinyurl.com/3qfup84 (user: guest, password: miroc)
Concluding remarks
MME is only one of the possible model ensembles
Extensive use of combined various ensembles as well as a
hierarchical modeling approach are desirable for advancing
our understanding of the cloud feedbacks
Hierarchical modeling
Zhang & Bretherson 2008
Medeiros et al. 2008
Brient & Bony 2012
Multi Model Ensemble
Multi Physics Ensemble
Gettelman et al. 2011
Watanabe et al. 2012
Perturbed Physics Ensemble
Collins et al. 2010
Yokohata et al. 2010
Sunderson et al. 2010
Klocke et al. 2012
Shiogama et al. 2012

Papers available at http://tinyurl.com/3qfup84 (user: guest, password: miroc)
Summary
Toward understanding sources of model’s “uncertainty”
in terms of the climate sensitivity & climate feedbacks,
We have shown thru the analysis of MIROC3.2/MIROC5/CMIP3
(& CMIP5 later) an importance of:
identifying timescales of cloud response/feedback
distinguishing robust/fragile parts associated w/
cloud response/feedback
Extensive use of various ensembles based on MIROC reveals
a connection between apparently independent
phenomena (ENSO & climate sensitivity)
attribution of inconsistency among GCMs to
individual processes/interactions